Comprehensive models to estimate the Isobaric heat capacity of deep eutectic solvents based on Machine learning algorithms

Abstract

Deep eutectic solvents (DESs) have recently gained significant attention due to their sustainable and environmentally friendly properties. Acquiring exact knowledge regarding the isobaric heat capacity (C_p) of DESs is essential for energy-related processes, in which these green solvents are utilized. Hence, this study deals with the development of comprehensive models for estimating the C_p of DESs. To reach this target, 682 experimental data, encompassing the C_p of 36 different DESs over widespread ranges of pressure and temperature, were assembled from the literature. The foregoing data were employed to establish new models based on four machine learning techniques, including Gaussian process method (GPM), Adaptive neuro-fuzzy inference system (ANFIS), Radial basis function neural network (RBF-NN) and Multilayer perceptron neural network (MLP-NN). The evaluations performed based on the statistical indices and graphical tools demonstrated that although all suggested models present excellent estimations, that designed based on the MLP-NN method yields the highest accuracy with Mean absolute percentage error (MAPE) and coefficient of determination (R²) values being 0.4% and 99.93, respectively, for the validation data. A comparison between the results of the literature correlations and those of novel models confirmed the obvious superiority of the latter. Moreover, the proposed models properly described the C_p variations of different DESs versus pressure and temperature. The order of importance of various factors in controlling the C_p of DESs was also determined based on the sensitivity analysis. Eventually, the intelligent models showed excellent performance in estimating the C_p of unseen DESs.